S100A8 and S100A9 are novel nuclear factor kappa B target genes during malignant progression of murine and human liver carcinogenesis


  • Potential conflict of interest: Nothing to report.


The nuclear factor-kappaB (NF-κB) signaling pathway has been recently shown to participate in inflammation-induced cancer progression. Here, we describe a detailed analysis of the NF-κB–dependent gene regulatory network in the well-established Mdr2 knockout mouse model of inflammation-associated liver carcinogenesis. Expression profiling of NF-κB–deficient and NF-κB–proficient hepatocellular carcinoma (HCC) revealed a comprehensive list of known and novel putative NF-κB target genes, including S100a8 and S100a9. We detected increased co-expression of S100A8 and S100A9 proteins in mouse HCC cells, in human HCC tissue, and in the HCC cell line Hep3B on ectopic RelA expression. Finally, we found a synergistic function for S100A8 and S100A9 in Hep3B cells resulting in a significant induction of reactive oxygen species (ROS), accompanied by enhanced cell survival. Conclusion: We identified S100A8 and S100A9 as novel NF-κB target genes in HCC cells during inflammation-associated liver carcinogenesis and provide experimental evidence that increased co-expression of both proteins supports malignant progression by activation of ROS-dependent signaling pathways and protection from cell death. (HEPATOLOGY 2009.)

Hepatocellular carcinoma (HCC) is the most frequent type of liver cancer and one of the most prevalent causes of cancer mortality worldwide. These tumors arise at sites of chronic liver injury, inflammation, and hepatocyte proliferation provoked by several causes such as chronic hepatitis B and C viral infection, chronic alcohol consumption, and aflatoxin B1–contaminated food.1, 2 Despite remarkable improvements in diagnosis, only limited therapeutic options exist, most of them with minimal clinical benefit.2 Moreover, there is only a rudimental understanding of the molecular, cellular, and environmental mechanisms that drive disease pathogenesis. In the past, numerous genetically modified mouse models have been established and extensively used to elucidate critical oncogenic pathways in the pathogenesis of HCC. For example, homozygous disruption of the multidrug resistance 2 (Mdr2) gene, which encodes for a P-glycoprotein and is responsible for the transport of phosphatidylcholine into the bile, leads to the development of cholestatic hepatitis followed by dysplasia, dysplastic nodules, and carcinoma.3 Thus, this mouse model represents a prototype of inflammation-associated cancer. Hepatitis and cancer progression were monitored in Mdr2−/− mice, showing the inflammatory process to trigger activation of the transcription factor nuclear factor kappaB (NF-κB) in hepatocytes through up-regulation of tumor necrosis factor-alpha (TNF-α) in adjacent endothelial and inflammatory cells.4 Besides the originally studied role of NF-κB in the context of adaptive and innate immune responses,5 experimental data demonstrated NF-κB to regulate cell proliferation, survival, and cell adhesion, and is constitutively active in different types of human cancer.6 In this regard, NF-κB activity was also found in hepatocytes, where it regulates liver cell proliferation and survival during development, regeneration, and neoplastic transformation.6, 7 Under normal conditions, the various Rel/NF-κB family members, RelA, RelB, c-Rel, NF-κB1, and NF-κB2 are sequestered in the cytoplasm because of physical interaction with inhibitors of NF-κB (IκBs).8, 9 Viral infection, DNA damage, pro-inflammatory cytokines, and stress stimulate the IκB kinase (IKK) complex, which comprises two catalytic subunits (IKKα and IKKβ) and a scaffold component (IKKγ). The IKK complex promotes NF-κB activation through phosphorylation-induced ubiquitination of IκBs and their subsequent degradation by the 26S proteasome, thereby enabling shuttling of NF-κB to the nucleus and binding to a common sequence motif, the κB site.9, 10 Notably, in the Mdr2−/− mouse tumor model, suppression of NF-κB activity in hepatocytes by an inducible transgene expressing an IκB-superrepressor (IκB-SR), a nondegradable IκB, had no effect on the course of hepatitis and on early phases of hepatocyte transformation, suggesting that NF-κB activity in hepatitis-induced HCC is dispensable for the early stage of tumor promotion.4 However, NF-κB inhibition at later stages resulted in apoptosis of precancerous hepatocytes and inefficient progression to HCC, supporting a critical role for the NF-κB–dependent genetic program at later stages of inflammation-associated hepatocarcinogenesis. Activation of NF-κB exerting its prosurvival function on transformed epithelial cells against pro-apoptotic cytokines secreted by the host immune system was also described in a mouse model of colitis-associated cancer using conditional ablation of IKKβ in enterocytes.11 However, the precise role of NF-κB signaling during liver carcinogenesis is still under debate as opposing effects have been described for different mouse tumor model systems. Whereas in Mdr2−/− mice NF-κB inhibition leads to significant reduction of HCC incidence, hepatocyte-specific ablation of its upstream kinases and thereby inactivation of NF-κB promotes tumorigenesis.4, 12, 13

In the current study, we performed a functional genomic approach with HCCs derived from Mdr2−/− (KO) and IκB-SR Mdr2−/− mice (DM) and identified S100A8 and S100A9 as novel NF-κB target genes that support malignant progression of HCC cells by activation of reactive oxygen species (ROS) and protection against apoptosis.


cDNA, complementary DNA; CM-H2DCFDA, 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate, acetyl ester; DM, IkappaB-SR Mdr2−/−; EMSA, electrophoretic mobility shift assay; HCC, hepatocellular carcinoma; HNE, 4-hydroxy-2-nonenal; IκB, inhibitor of nuclear factor kappa B; IκB-SR, inhibitor of nuclear factor kappa B superrepressor; IKK, inhibitor of nuclear factor kappa B kinase; KO, Mdr2−/−; MAPK, mitogen-activated protein kinase; Mdr2, multidrug resistance 2; NF-κB, nuclear factor-kappaB; ROS, reactive oxygen species; RQ-PCR, quantitative real-time polymerase chain reaction; SEM, standard error of the mean; TNF-α, tumor necrosis factor-alpha.

Materials and Methods


Mice with homozygous disruption of the Mdr2 gene were bred to transgenic mice (IκB-SR) carrying two transgenes: the IκB-superrepressor controlled by a tetracycline-regulated promoter and the tetracycline-transactivator under control of the hepatocyte-specific C/EBPβ promoter (TALAP1). All animal experiments were performed as described elsewhere.4

RNA Preparation from Normal Liver and HCCs.

Total RNA extraction was performed according to the manufacturer's instructions using TRIZOL Reagent (Invitrogen, Karlsruhe, Germany). Quality check of the isolated RNA, in vitro amplification, and sample labeling was performed as described previously.14, 15

Quantitative Real-Time Polymerase Chain Reaction Analysis.

Quantitative real-time polymerase chain reaction (RQ-PCR) analysis was performed as described previously.16 Primers used for RQ-PCR analysis are listed in Supporting Table 2. Target gene cycle of threshold values were normalized to the corresponding cycle of threshold values of Hprt (for mouse samples) or LAMINB1 (for human samples) using the change in cycle of threshold method.

Immunohistochemistry Analysis.

Immunohistochemistry staining was performed on mouse liver sections from KO and DM mice and on a tissue microarray containing human normal liver, peritumorous, and tumor tissue17 with the Immunodetection Kit (Vector Laboratories; Burlingame, CA) according to manufacturer's instructions. Primary and secondary antibodies used are listed in Supporting Table 3.

Transfection of HCC Cell Lines.

The HCC cell lines Hep3B and HuH-7 were transiently transfected with expression plasmids using FuGENE HD Transfection Reagent according to the manufacturer's instructions (Roche, Basel, Switzerland). Following expression plasmids were used: pRSV-0, pRSV-RelA, pcDNA3.1/Myc-His (C) (Invitrogen), pcDNA3.1-S100A8MycHis (encodes a S100A8-Myc-Tag/6xHistidin-Tag fusion protein), and pcDNA3.1-S100A9Flag (encodes a S100A9-Flag-Tag fusion protein).18 Cells were co-transfected with an expression plasmid (pMSCV-EGFP) encoding an enhanced green fluorescent protein under control of a CMV-LTR control element. The transfection efficiency determined by fluorescence-activated cell sorting analysis reached 70% for Hep3B and 35% for HuH-7 cells.

Electrophoretic Mobility Shift Assays.

Electrophoretic mobility shift assays (EMSA) were performed with biotinylated oligonucleotides as listed in Supporting Table 4, using the LightShift Chemiluminescent EMSA Kit (Pierce, Rockford, IL). For bandshifts using Oct oligonucleotides, binding buffer was supplemented with 1 μg poly(dI-dC), 12.5% (vol/vol) glycerol, 50 mM KCl, and 0.5 mM ethylenediamine tetraacetic acid. Bandshifts with kappaB oligonucleotides were performed with binding buffer supplemented with 1 μg poly(dI-dC), 12.5% (vol/vol) glycerol, 100 mM KCl, 0.5 mM ethylenediamine tetraacetic acid, and 0.1% (vol/vol) Triton X-100. For the detection of biotin-labeled nucleic acids, the Chemiluminescent Nucleic Acid Detection Module was used according to the manufacturer's instructions (Pierce).

ROS Activity Measurement.

ROS was detected by 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate, acetyl ester (CM-H2DCFDA) (Invitrogen Molecular Probes, Eugene, IL). Transfected Hep3B cells (48 hours after transfection) were stimulated with 40 ng/mL TNF-α (Sigma-Aldrich, Munich, Germany) for 1 hour, followed by loading with 20 μM CM-H2DCFDA for 30 minutes at 37°C. Additional treatment with 1 mM H2O2 for 5 minutes served as a positive control. After loading, cells were trypsinized and diluted in phosphate-buffered saline for flow cytometry analysis. Mean fluorescence intensities were used as measures for ROS levels determined by fluorescence-activated cell sorting analysis using CellQuest Pro (BD Biosciences, Heidelberg, Germany).

Microarray Experiments, Statistical Analysis, Preparation of Nuclear Extracts, Cultivation of Cells, Western Blot Analysis, Annexin-V Staining, Small Interfering RNA Experiments, Tissue Preparation, Immunofluorescence Analysis.

See Supporting Experimental Procedures.


Identification of NF-κB–Regulated Genes in the Mdr2−/− Mouse Tumor Model.

To identify tumor-associated genes, whose expression critically depends on hepatocellular NF-κB function, we isolated total RNA from specimens with the same tumor grade of four Mdr2−/− (KO) and six IκB-SR Mdr2−/− (DM) mice4 and performed global gene expression analysis. The expression level of the IκB-SR protein in livers of DM mice was determined by western blot analysis to confirm conditions of efficient inhibition of NF-κB function (Supporting Fig. 1). We found 367 differentially expressed and annotated genes, of which 320 were up-regulated and 47 were down-regulated in KO compared with DM HCC samples (Fig. 1 and Supporting Table 1, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=dxstdssgkykswzk&acc=GSE13599). Gene clustering according to functional annotation indicated distinct physiological and pathological features, such as inflammatory disease, immune response, cellular growth, and proliferation, which are well known to be regulated by NF-κB signaling (Fig. 1). Furthermore, at least 26 of the 367 differentially expressed genes harbored a conserved NF-κB binding site within the promoter region, which resulted in a significant association of putative NF-κB dependency and differential expression as determined by a chi-square test (P = 0.014). In addition to previously identified NF-κB target genes, such as Cd74, Cxcl9, Icam1, Spp1, Tapbp, and Vcam1 (http://people.bu.edu/gilmore/nf-kb/target/index.html and http://bioinfo.lifl.fr/NF-KB/), 12 genes present in our list failed the binding motif alignment by one mismatch in either mouse or human, though being well-known NF-κB targets (for example, Cd38 and Gbp1). Ninety-six percent of the differentially expressed genes sharing a NF-κB binding site were up-regulated in KO versus DM HCCs, providing experimental evidence that the NF-κB response pathway was constitutively active. RQ-PCR analyis of the well-known NF-κB target genes Cd38 and Icam1 confirmed different transcript levels in KO and DM HCC samples (Fig. 2A). A similar expression pattern was found for S100a8 and S100a9, two differentially expressed genes present in the cluster of inflammatory disease (Fig. 2A and Supporting Fig. 2), suggesting that S100a8 and S100a9 represent novel candidate genes of NF-κB during inflammation-associated carcinogenesis.

Figure 1.

Heat map of differentially expressed genes between KO and DM HCCs. Differentially expressed genes derived from the global gene expression analysis with cDNA from KO versus DM HCCs were clustered hierarchically. Genes were grouped according to their functional annotation as assigned by Ingenuity Systems. Indicated expression values are normalized log2 ratios of each individual sample against wild-type liver. Only genes with functional assignment according to Ingenuity Systems are shown.

Figure 2.

Expression of S100a8 and S100a9 depends on active NF-κB during liver carcinogenesis. (A) Relative transcript levels of well-known NF-κB target genes (Cd38 and Icam1) and potential new candidate genes (S100a8 and S100a9) were determined by RQ-PCR analysis comparing KO and DM HCCs, normalized to the expression in WT liver. Bars represent mean expression values of four KO and six DM HCCs ± standard error of the mean (SEM). (B) Transcript levels of S100a8 and S100a9 during liver carcinogenesis were compared by RQ-PCR analysis using cDNA of liver tissue from 3-month-old, 7-month-old, and 10-month-old animals and of HCC tissue derived from 16-month-old to 18-month-old KO and DM mice. The expression levels were normalized to the expression of S100a8 and S100a9 in WT liver. Bars represent mean expression values of at least three different samples ± SEM.

NF-κB-Dependent Expression of S100a8 and S100a9 in Hepatocellular Carcinoma Cells.

To determine the onset of NF-κB–dependent S100a8 and S100a9 transcription during HCC development and progression, we performed RQ-PCR analysis with complementary DNA (cDNA) derived from premalignant and malignant liver samples of KO and DM mice. In line with the global gene expression data, we measured elevated expression of S100a8 and S100a9 in KO HCCs but found no major alteration in liver samples of 3-month-old, 7-month-old, or 10-month-old animals (Fig. 2B). Accordingly, immunohistochemical analysis of KO liver tissue sections revealed intracellular S100a8 and S100a9 protein staining in both hepatocellular carcinoma cells and stromal cells. Although similar amounts of S100a8 and S100a9-positive stromal cells were present in KO versus DM HCC samples (Supporting Fig. 3), expression of both proteins was not found in HCC cells of DM mice (Fig. 3, Supporting Fig. 4), suggesting that elevated transcript levels are most likely attributable to NF-κB–dependent transcription in HCC cells.

Figure 3.

S100a9 protein expression in KO and DM liver sections. S100a9 protein expression in liver tissue sections from 3-month-old, 7-month-old, and 10-month-old animals and of HCC tissue derived from KO and DM mice was monitored by immunohistochemistry analysis using an S100a9-specific antibody. Signals for S100a9 were detected in infiltrating cells on liver sections from both KO and DM animals during all stages of liver carcinogenesis, whereas S100a9 protein expression in hepatocytes was restricted to KO HCCs (arrows in the inlet). The staining of DM HCCs only shows expression of S100a9 in the inflammatory cluster. Representative images of n = 3 animals of each genotype are shown with red staining for S100a9 protein and counterstaining with hematoxylin. Scale bar = 100 μm.

S100A8 and S100A9 Co-expression in Human HCC Tissue Samples.

To investigate whether S100A8 and S100A9 are also co-expressed in human HCCs, we stained sections on a tissue microarray covering 68 HCC samples.17 Indeed, we found co-expression of both proteins in HCC cells and in cells of the tumor stroma (Spearman rank order correlation = 0.324, P = 0.005) (Fig. 4A). In addition, co-expression of S100A8 and S100A9 protein in HCC cells of human HCCs was verified by co-immunofluorescence analysis (Supporting Fig. 5). S100A8 and S100A9 belong to a large family of S100 proteins encoded by genes that are tightly clustered on human chromosome 1q21. Recent studies using comparative genomic hybridization analysis showed a prominent amplification on chromosome 1q in human HCCs.19 To rule out the possibility that a gene dosage effect is responsible for elevated S100A8 and S100A9 expression, we quantified the transcript levels for both genes in human HCC samples with or without 1q amplification. Indeed, RQ-PCR analysis showed similar mean expression values for all HCCs independent of the genomic status (Fig. 4B), suggesting that regulation of gene transcription, such as by NF-κB signaling, rather than a gene dosage effect is implicated in enhanced S100A8 and S100A9 transcript levels in HCC cells.

Figure 4.

S100A8 and S100A9 are significantly coexpressed in human HCC. (A) S100A8 and S100A9 protein expression was determined by immunohistochemistry analysis using S100A8-specific and S100A9-specific antibodies on a tissue microarray containing normal liver samples, peritumorous tissue, and HCCs. Representative images of normal liver, peritumorous, and HCC tissue are shown (staining in red and counterstaining with hematoxylin). (B) RQ-PCR analysis was used to quantify transcription levels of S100A8 and S100A9, normalized to LAMINB1 (LMNB1), in human HCCs with balanced (n = 10) or amplified 1q chromosomal locus (n = 20). Box blots show mean expression values for S100A8 and S100A9 ± SEM.

NF-κB Induces S100A8 and S100A9 Transcription in Human HCC Cell Lines.

To unravel the molecular mechanism of induced S100A8 and S100A9 expression in HCC cells, we transfected the HCC cell line Hep3B with an expression plasmid for the NF-κB subunit RelA and quantified basal and inducible transcription of both genes by RQ-PCR analysis. We detected basal S100A9 transcription in mock transfected Hep3B cells and a 12-fold increase by ectopic RelA expression (Fig. 5A). In contrast, only very low basal S100A8 transcription was detectable; however, ectopic RelA expression resulted in a tremendous increase of S100A8 transcript levels compared with mock controls (Fig. 5A). Furthermore, stimulation of different human HCC cell lines (Hep3B, PLC/PRF/5, and HuH-7) with TNF-α, a potent inducer of NF-κB signaling, induced transcription of S100A9, but not S100A8 (data not shown). These data imply that NF-κB activity in HCC cells is implicated in elevated S100A8 and S100A9 co-expression. In addition, they suggest that S100A9 is a direct NF-κB target gene, whereas regulation of S100A8 requires sustained NF-κB activation or synergistic function of other transcriptional regulators. In line with this assumption, detailed inspection of S100A8 and S100A9 gene promoters revealed a conserved NF-κB binding motif for both human and mouse S100A9. However, the human S100A8 promoter contains only a κB-like motif, which is not conserved in the mouse promoter (Fig. 5B). To prove direct NF-κB binding to the human S100A9 promoter upon ectopic RelA expression, we performed EMSA with biotinylated oligonucleotides including the predicted NF-κB binding site in the human S100A9 promoter. An inducible bandshift was detected with nuclear extracts of RelA-transfected Hep3B cells compared with mock controls with oligonucleotides containing either a conserved κB-site or the κB motif of the human S100A9 promoter. Introduction of specific mutations into the κB-site of the human S100A9 promoter led to a complete loss of complex formation. Importantly, no bandshift was seen using the oligonucleotides containing the κB-like site predicted in the human S100A8 promoter (Fig. 5C). Finally, transfection of Hep3B cells using several small interfering RNA oligonucleotides that repress either NF-κB1 or RelA protein levels showed significant down-regulation of basal S100A9 transcription (Supporting Fig. 6). In summary, these in vitro data confirm our results obtained from the global gene expression analysis that S100a8 and S100a9 are regulated by NF-κB and S100A9 represents a novel direct NF-κB target gene in HCC cells.

Figure 5.

S100A9 is a direct NF-κB target gene in HCC cells. (A) Ectopic expression of RelA in the human HCC cell line Hep3B promoted strong induction of S100A8 and S100A9 gene expression. RQ-PCR analysis was used to determine relative transcript levels for S100A8 and S100A9 in cDNA derived from mock versus RelA-transfected Hep3B cells 48 hours after transfection. Bars represent mean expression values of n = 3 independent transfections ± SEM. (B) Detailed analysis of human S100A8 and S100A9 promoters revealed a conserved κB binding site in the S100A9 promoter and a κB-like site in the S100A8 promoter. S100A9_mut defines the oligonucleotide in which the κB binding site in the S100A9 promoter is mutated. (C) EMSA with nuclear extracts of mock versus RelA-transfected Hep3B cells showed induced DNA-binding activity with the conserved control κB oligonucleotide (KappaB) and the oligonucleotide sharing the κB binding site in the S100A9 promoter (S100A9). Specific mutation of the κB binding site in the S100A9 promoter led to a complete loss of complex formation (S100A9_mut). EMSA with an oligonucleotide containing the κB-like site in the S100A8 promoter did not show any bandshift (S100A8). Bandshifts with an Oct oligonucleotide served as control for quality and quantity of the nuclear extracts (Oct).

S100A8 and S100A9 Expression in Hepatocellular Carcinoma Cells Activates ROS-Dependent Signaling Pathways.

There is evidence that enhanced production of ROS is a common response to all described HCC-inducing causes. Because S100A8 and S100A9 overexpression in HaCaT keratinocytes (immortalized human keratinocytes) increases nicotinamide adenine dinucleotide phosphate (NADPH) oxidase activity and enhances ROS levels,20 we asked whether elevated S100A8 and S100A9 protein levels induce ROS production also in HCC cells. Hep3B cells were transfected with S100A8 and S100A9 expression plasmids, and ROS levels were measured by DCFDA fluorescence staining using flow cytometry analysis. No or only mild differences in ROS levels were detected between Hep3B cells with ectopic expression of either S100A8 or S100A9 alone and mock controls (Fig. 6A). However, ROS production was significantly elevated in S100A8/S100A9-transfected Hep3B cells compared with mock controls (P = 0.0124), suggesting that the presence and synergistic function of both proteins is critical for enhanced ROS production. This effect was even more pronounced when transfected cells were treated with TNF-α (Supporting Fig. 7). Next, we determined the level of 4-hydroxy-2-nonenal (HNE) protein adducts, a marker of oxidative-stress induced damage,21 in normal liver and HCC samples by immunohistochemistry analysis. Indeed, hepatocytes with strong staining for HNE-protein adducts were easily found on tumor sections of KO but not DM mice (Fig. 6B), demonstrating a positive correlation between elevated ROS levels and induced S100a8/S100a9 expression also in vivo. To unravel the signaling pathways mediated by enhanced ROS levels, we studied the regulation of p38 mitogen-activated protein kinase (MAPK), a well-known downstream target, and detected less p38 MAPK phosphorylation in S100A8 and S100A9 expressing cells treated with TNF-α in comparison with mock controls (Supporting Fig. 8). Phosphorylation of ATF-2, a downstream target of the p38 MAPK pathway, was also impaired in TNF-α–treated S100A8/S100A9-expressing cells compared with mock controls (Supporting Fig. 8).

Figure 6.

Ectopic S100A8 and S100A9 expression enhances ROS activity in HCC cells. (A) Hep3B cells were transfected with expression plasmids for human S100A8 and S100A9 (S100A8/S100A9) either separately or in combination and empty vector (mock) as control. Expression of S100A8 and S100A9 protein in transfected cells was determined by western blot analysis using anti-myc (S100A8-Myc) or anti-Flag (S100A9-Flag) antibodies. Actin served as a loading control. ROS activity in mock (green line), S100A8 (blue line), S100A9 (purple line), or S100A8/S100A9 (red line) transfected Hep3B cells was determined by flow cytometry analysis using CM-H2DCFDA as substrate. Treatment of the cells with H2O2 for 5 minutes served as a positive control for maximal ROS production. (B) HNE-protein, adducts S100a8, and S100a9 proteins were detected by immunohistochemistry analysis using specific antibodies. Representative images for HNE, S100a8, and S100a9 stainings are shown in red (HNE-protein adducts) or brown (S100a8 and S100a9) for normal liver and HCCs of KO and DM mice and counterstaining with hematoxylin. Scale bar = 100 μm. Inlets represent higher magnification.

Ectopic S100A8/S100A9 Expression Protects Hepatocellular Carcinoma Cells from Induced Apoptosis.

Because NF-κB is a major regulator of cell survival, we asked whether the newly identified NF-κB targets S100A8 and S100A9 are also involved in this process. To this end, we measured cell death in Hep3B as well as HuH-7 cells with ectopic expression of S100A8/S100A9 in comparison with mock controls by annexin-V staining (Fig. 7 and Supporting Fig. 9). In Hep3B cells, a strong dose-dependent increase in annexin-V staining was measured for mock transfected cells after TNF-α treatment, demonstrating a significant induction of apoptosis (P = 0.0154). In contrast, the apoptosis rate in cells with ectopic S100A8/S100A9 expression was almost unchanged after TNF-α treatment (P = 0.2002). Similar results were observed in HuH-7 cells in which ectopic S100A8/S100A9 expression protected from basal as well as TNF-α–induced apoptosis (Supporting Fig. 9).

Figure 7.

Ectopic expression of S100A8 and S100A9 protects from TNF-α–induced apoptosis. Hep3B cells were transfected with expression plasmids for human S100A8 and S100A9 (S100A8/A9) or empty vector as control. Transfected cells were treated with 20 nM or 50 nM human TNF-α 24 hours after transfection. Annexin-V staining was performed 18 hours after TNF-α treatment. (A) Histograms show annexin-V staining of mock and S100A8/S100A9-transfected cells ± TNF-α treatment measured by flow cytometry analysis. Unstained cells (gray filling) served as negative control. (B) The percentage of annexin-V–positive cells determined by flow cytometry analysis in transfected cells ± TNF-α treatment was quantified. Bars represent mean percentage of annexin-V–positive cells of n = 3 independent experiments ± SEM.


Recent findings have implicated constitutive activation of the transcription factor NF-κB as a key event during neoplastic progression of the liver after viral infection, growth factor stimulation, and inflammation.7, 22, 23 Nevertheless, the function of NF-κB in liver tumorigenesis appeared to be much more complex, because loss of NF-κB function by ectopic expression of an IκB superrepressor inhibits tumor development, whereas hepatocyte-specific ablation of upstream kinases promotes tumorigenesis.4, 13, 24 Meanwhile, more detailed analyses of the different mouse models inducing liver cancer support the assumption that differences in tumor model systems and the temporal pattern of NF-κB activation might explain the contradictory findings during liver carcinogenesis.25

The Mdr2−/− mouse model is a well-established tumor model of inflammation-associated liver carcinogenesis. Recent data demonstrated a crucial role for NF-κB in hepatocytes during malignant progression by supporting survival of precancerous cells.4 Here, we used expression profiling to identify the NF-κB–dependent genetic program that drives tumorigenesis in this mouse model. We identified a comprehensive list of differentially expressed genes between KO and DM HCC samples, some of which represent well-known NF-κB target genes (Fig. 8). In addition, we found numerous novel putative NF-κB targets and studied the regulation and function of two candidates, S100a8 and S100a9, in more detail. Originally, S100A8 and S100A9 were discovered as an immunogenic complex expressed and secreted by neutrophils with potent antimicrobial properties.26 Meanwhile, a large body of experimental evidence demonstrates that S100A8 and S100A9 are pro-inflammatory mediators released by cells of myeloid origin in response to cell damage, infection, or inflammation, and function as pro-inflammatory danger signals.27, 28 Accordingly, elevated serum levels of S100A8 and S100A9 proteins are hallmarks of inflammatory disorders and recurrent infections, and increased expression of both proteins was evident in tumor-infiltrating immune cells of many epithelial malignancies.28, 29 In line with these observations, we detected S100A8 and S100A9 proteins in inflammatory cells infiltrating precancerous lesions and HCCs of the Mdr2−/− mouse model and of human patients. Interestingly, tumor-induced up-regulation of S100a8 and S100a9 proteins was found in myeloid precursor cells and contributes to the recruitment and accumulation of myeloid-derived suppressor cells at sites of tumor formation.30, 31 Myeloid-derived suppressor cells facilitate carcinogenesis and tumor progression by suppression of T cell and NK cell activation, thereby inhibiting antitumor immunity.32 Consequently, S100a9-deficient mice mounted a potent antitumor immune response leading to efficient rejection of implanted tumors.30

Figure 8.

Model of NF-κB–dependent gene interaction network in HCC development. Gene interaction network was generated by applying Ingenuity Pathway Analysis software (Ingeuity Systems), based on the identified gene signature. Differential gene expression is gradually depicted as red filling for up-regulation and green filling for down-regulation in KO versus DM HCCs. NF-κB–dependent expression of S100A8 and S100A9 induces ROS production and thereby is contributing to tumor promotion.

Importantly, we found no significant difference in the number of S100a8-positive or S100a9-positive stromal cells in the livers of KO and DM mice, suggesting that NF-κB–dependent alterations in transcript levels result from changes in hepatocytes. Indeed, we detected both proteins in HCC cells of KO mice as well as in human HCC specimens, confirming the observation that S100A8 and S100A9 proteins are produced not only by immune cells but also by epithelial tumor cells.27, 28

In the past, in vitro and in vivo model systems, including inflammation-associated colon carcinogenesis, supported a direct link between S100A8/S100A9 expression and activation of NF-κB signaling.33, 34 An intriguing feature of our study was the finding that S100A8 and S100A9 are not only inducers of NF-κB signaling, but also NF-κB-dependent target genes, supporting the existence of a feed-forward loop promoting inflammation and neoplastic transformation. Accordingly, we provide experimental evidence that S100A9 is a direct NF-κB target gene and could identify a functional κB-site in its proximal promoter region. We found S100A8 and S100A9 coexpression in HCC cells of KO mice in human HCC samples as well as strong up-regulation of S100A8 transcripts on ectopic RelA expression in Hep3B cells. However, we could not identify a functional NF-κB binding site in its proximal promoter region, supporting the hypothesis that S100A8 is either an indirect NF-κB target gene or requires synergistic action with other transcription factors. Recently, Cheng and colleagues demonstrated that Stat3 critically contributes to S100a8 and S100a9 expression in mouse myeloid precursor cells and demonstrated direct Stat3 binding at promoter regions of both genes.30 Stat3 represents an oncogene that is persistently activated in many human tumors, including HCCs, and its activation provides an important link between inflammation and cancer.35 Although several reports have already documented physical and functional interactions between Stat3 and NF-κB,36 further work is required to delineate a potential role of Stat3 for induced S100A8 expression in hepatocellular tumor cells.

Recently, S100A8 and S100A9 were found to induce ROS production in HaCaT keratinocytes by increasing NADPH oxidase activity.20 In line with this observation, we found that S100A8 and S100A9 co-expression enhances ROS levels in HCC cells that were even more pronounced in TNF-α–stimulated cells. Numerous studies demonstrate a causal link between ROS and molecular mechanisms of liver carcinogenesis and defined ROS as a common denominator of signaling pathways induced by different HCC-promoting agents.37 Oxidative stress exerts many pro-tumorigenic effects, such as altered gene expression, enhanced proliferation, high DNA mutation rates, as well as genomic instability.38 Importantly, administration of antioxidants and selenium compounds to Mdr2−/− mice reduced tumor incidence and growth by inhibition of mediators that control inflammation and response to oxidative stress.39 In addition to induced ROS production in cells with ectopic S100A8 and S100A9 expression, we found reduced phosphorylation of the p38 MAPK in comparison with mock controls. These data are in line with the observation that p38 MAPK function was down-regulated through selective dephosphorylation after oxidative stress because of rapid stress-induced activation of the phosphatase MKP-1 in vivo.40 However, we cannot exclude the possibility that intracellular S100A8 and S100A9 interfere with p38 MAPK activity and thereby cause ROS production, as has been demonstrated in liver-specific p38α knockout mice.38 Detailed analysis will be required to unravel the exact molecular mechanism; however, p38α has already been shown to suppress normal and cancer cell proliferation by antagonizing the c-Jun N-terminal kinase/cJun pathway.41 In addition, we demonstrated that HCC cell lines with ectopic expression of S100A8 and S100A9 were almost resistant against TNF-α–induced apoptosis. This observation supports the assumption that NF-κB and its downstream targets protect precancerous cells from cytokine-induced cell death in vivo.

In summary, we identified S100A8 and S100A9 as novel NF-κB target genes in mouse and human HCCs. We assume that their co-expression promotes malignant progression by induction of ROS production, down-regulation of p38 MAPK signaling, and support of cell survival. Our findings raise the intriguing possibility that pharmacological inhibition of their function during liver carcinogenesis may selectively eradicate malignant liver cells without affecting normal liver homeostasis.


We thank Moritz Küblbeck, Ingeborg Vogt, Johannes Müggenburg, and especially Sibylle Teurich for excellent technical assistance. We thank Dr. Tobias Dick for support in ROS measurements, Dr. Rüdiger Arnold (DKFZ, Heidelberg), and Dr. Thomas Wirth (University, Ulm) for providing TNF-α and RSV-RelA, respectively.